which metric identifies patches with the nieghborhood rule?

This feels like this should be the very simplest thing one can do in fragstats but its not working! Or more likely I don't understand it properly.

I want to identify patches using the 8-neighborhood rule, and I assumed that "patch area" with the 8-nieghborhood rule selected would identify patches of a given class that comprise at minimum a contiguous area of at least the focal pixel with the 8-pixel neighborhood. However, the output includes isolated individual pixels of that class. Is there somewhere I need to tell it to use the 8 neighborhood rule to define a patch?

Alternatively, if you could point me to resources that explain where the 4/8 neighborhood rule is used that would help. (Because in the description for PI Area it sounds like it should use it). I could query out by patch size but then would still need to deal with shape, since a long linear line of pixels would also not meet our criteria.

Thanks for any info!

Hello kls430,

All patches are identified according to the selected neighborhood rule, metrics are calculated afterwards using patch characteristics.
4-neighbor rule means that 2 cells of the same class are part of the same patch if they have a common side.
8-neighbor rules means that 2 cells of the same class are part of the same patch if they have a common side or if they touch diagonally, i.e. toughing corners.
If you consider a cell in a grid surrounded by its 8 adjacent neighbors, in the 4-neighbor rule case there are just 4 possible neighbors (up, down, left, right) if the class id matches while in the 8-neighbor case there are 8 possible neighbors if the class id matches.
The patch is constructed by adding neighbors that match the ID and the neighborhood rule, then the construction process moves iteratively through the added neighbors until there is no more to add. It then moves to the next viable untouched cell.

I hope that clarifies it. We also strongly recommend reading the documentation.

Best regards,
Eduard

I did read the documentation. From that and from your description, it is not clear to me why Patch Area with is identifying single, individual pixels that do not have a neighbor in the same class, as a patch.

Hello kls430,

The identification of patches happens at the beginning of the analysis and by definition, the smallest possible patch is a patch of 1 cell. And it cannot have a neighbor of the same class because that would make it a bigger patch (neighborhood rules apply here).

Metrics are then calculated against the list of identified patches regardless of the number of cells they consist of. That is exactly as it should be, we don't see any problem here.

Best regards,
Eduard